Capture and Exploration of Sample Quality Data to Inform and Improve the Management of a Screening Collection

Abstract

A new approach to the storage, processing, and interrogation of the quality data for screening samples has improved analytical throughput and confidence and enhanced the opportunities for learning from the accumulating records. The approach has entailed the design, development, and implementation of a database-oriented system, capturing information from the liquid chromatography-mass spectrometry capabilities used for assessing the integrity of samples in AstraZeneca's screening collection. A Web application has been developed to enable the visualization and interactive annotation of the analytical data, monitor the current sample queue, and report the throughput rate. Sample purity and identity are certified automatically on the chromatographic peaks of interest if predetermined thresholds are reached on key parameters. Using information extracted in parallel from the compound registration and container inventory databases, the chromatographic and spectroscopic profiles for each vessel are linked to the sample structures and storage histories. A search engine facilitates the direct comparison of results for multiple vessels of the same or similar compounds, for single vessels analyzed at different time points, or for vessels related by their origin or process flow. Access to this network of information has provided a deeper understanding of the multiple factors contributing to sample quality assurance.

Rapid Compound Integrity Assessment for High-Throughput Screening Hit Triaging

“…Data processing was performed using Analytical Data Reviewer (Virscidian, Cary, NC), which processed the raw instrument files, calculated the UV percent purity for the target chromatographic peak, and determined if the correct monoisotopic mass peaks were found in the mass spectra. 10 After automated processing by Analytical Data Reviewer, manual review was then performed for all samples with nongreen (<75% of confirmed purity) results. On average, about 40% of samples on each plate underwent manual review.…”

Toward Automated Interpretation of LC-MS Data for Quality Assurance of a Screening Collection

“…A diagram and description of the manual review process are provided in an earlier article. 8 Cross-validation was used for the initial assessment of learning algorithms and for the experiments on pruning methods. Smaller training sets are more likely to be biased by particular features of the data and therefore suffer from overfitting.…”

“…7 The overall quality of AstraZeneca's corporate compound collection is monitored by Compound Management (CM). The purity and structural identity of compounds entering the collection are established using reversed-phase high-performance chromatography coupled with MS with an additional CAD to measure concentration, 8 hereafter referred to as the LC-MS system. In addition, LC-MS activities are required to support the formation of screening collections for high-throughput screening (HTS) campaigns as well as the provision of "just in time" analysis to support later-stage screening projects.…”

“…As a result, 33.8% of the analyses performed were automatically flagged as "passed," with the remaining analyses reviewed by domain experts. 8 Despite the success of the autopass functionality, the interpretation of analytical data remains the bottleneck in the LC-MS workflow. The available archive of more than 700,000 analyses, many of which were annotated by LC-MS experts, was identified as an opportunity to apply a data-mining approach to further automate the classification of LC-MS results.…”

Maintaining a High-Quality Screening Collection: The GSK Experience

“…Published work covering chemical instability in DMSO has been limited, 4–8 and attempts to build a predictive model have been only partially successful. 9,10 This is not surprising, considering the range of chemistries involved.…”